Network survivability is a key concern in today's network, and will become increasingly important in future optical networks as they carry ever more traffic. Networks are also becoming more complex, with the requirement for increased functionality. Currently, there is a lack of understanding in the industry as to the exact relationship between the choice of network architecture and the meeting of a set availability objective. This paper analyses a number of long-haul network architectures from an unavailability point of view. The long-haul networks analyzed include: networks with diversity, networks with restoration capability, and networks with survivability. Derivations are given for each architecture; formulas for 2 and 4-fiber rings, and dual fed routing are new. A hypothetical reference connection (HRX) and its unavailability objectives are used as references. Networks with restoration capability and networks with survivability meet the proposed objective. Self-healing rings (both 2 and 4-fiber bidirectional line switched rings) and dual fed systems offer the highest level of survivability, by eliminating service impacts caused by cable cuts and equipment failures.>